Sepsis and Septic Shock
Sepsis is not a fever and it is not a lab value. It is infection plus organ failure, and the two things that change the outcome are the right antibiotic and the drainage, not the size of your fluid bolus.
The definitions that actually matter
Sepsis-3 defines sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection. Both halves are load bearing. No infection, no sepsis. No organ dysfunction, no sepsis either: that is an infection, and infections without organ failure are common and mostly fine.
Organ dysfunction means an acute rise in the SOFA score of 2 points or more from baseline, scored across respiratory (P/F ratio), coagulation (platelets), liver (bilirubin), cardiovascular (hypotension and pressor dose), neurologic (Glasgow Coma Scale) and renal (creatinine and urine output). Assume baseline is zero absent known chronic dysfunction. In practice: the patient who was fine yesterday and today has a creatinine of 2.1, platelets of 90, a new oxygen requirement or new confusion.
Septic shock
Septic shock needs two things, not one:
- A vasopressor requirement to hold a mean arterial pressure at or above 65 mmHg, and
- A lactate above 2 mmol/L, both persisting after adequate volume resuscitation.
That last clause is where people get sloppy. A hypotensive patient who has received nothing is not in septic shock yet, and a patient on norepinephrine with a normal lactate has vasoplegia but does not meet the definition. The distinction matters for prognosis and, unglamorously, for coding.
qSOFA, and why I do not want you screening with it
Quick SOFA is three items, one point each: respiratory rate 22 or higher, altered mentation, systolic pressure 100 or lower. Two or more is positive. It needs no labs, which is why it spread.
It is decent at telling you which already-identified infected patient will do badly. It is poor at finding sepsis early, because its sensitivity is low. By the time someone is tachypneic, confused and hypotensive you did not need a score. Validation work has repeatedly found it less sensitive than SIRS or a standard early warning score for catching infected patients who deteriorate, and the Surviving Sepsis authors now recommend against using it as a single screening tool. Screen with an early warning system and use qSOFA, if at all, to sharpen worry you already have.
What happened to SIRS
Two of four: temperature above 38 or below 36 degrees Celsius, heart rate above 90, respiratory rate above 20 or PaCO2 below 32, and white count above 12,000, below 4,000, or over 10 percent bands. It was abandoned for two reasons. It is wildly nonspecific: a marathon runner, a postoperative patient, someone in alcohol withdrawal and anyone who just got frightened all qualify, and a definition that captures half the medical floor defines nothing. It is also not sensitive enough, since a meaningful minority of patients with infection and real organ failure never meet two criteria, particularly the elderly, the immunosuppressed and anyone on a beta blocker who cannot mount a tachycardia. Sepsis-3 also retired severe sepsis, since sepsis by the new definition is already severe. You will still hear it, because coding moves slower than consensus panels.
The first hour
Two sets from separate sites, plus a culture of whatever you think the source is. Drawing first substantially raises your chance of an organism, and an organism is what lets you narrow later. But if access is taking ten minutes, or the lumbar puncture will take forty, give the drug. A missed organism is a problem. An hour of untreated septic shock is a bigger one.
Send it, repeat in 2 to 4 hours if above 2. It is not a pure measure of tissue hypoxia: catecholamines, impaired hepatic clearance, beta agonists and thiamine deficiency all raise it. It is still the most useful single number for how sick this person is, and the trend beats the value.
In septic shock, target one hour from recognition. In sepsis without shock you have a little room to think, and thinking is worth something. Do not let the drug wait behind a CT scanner or a verification queue.
Do not wait for the fluid to finish. If the patient is profoundly hypotensive, start norepinephrine while the fluid runs. Perfusion pressure is not earned by completing a checklist item in order.
About the 30 mL/kg
The recommendation is 30 mL/kg of crystalloid in the first 3 hours for sepsis-induced hypoperfusion. It is a reasonable starting point, roughly the volume patients received in the early goal-directed therapy era. It is not a physiologic law and not a number to chase into someone already wet. For a 70 kg patient it is 2.1 liters; for a 120 kg patient, actual body weight gives 3.6 liters and will drown people, so use ideal or adjusted weight in significant obesity. Give it in 250 to 500 mL aliquots, reassessing pressure, perfusion, work of breathing and oxygen requirement after each. When the patient stops responding, stop. Balanced crystalloid is the sensible default and the exceptions are on the IV fluids page.
Heart failure with reduced ejection fraction. Thirty mL/kg into an ejection fraction of 20 percent buys you pulmonary edema and an intubation. Give 250 mL, look at the patient, reach for norepinephrine early.
End-stage renal disease on dialysis. No escape valve. What you give stays until someone dialyzes it off.
Severe pulmonary hypertension or right ventricular failure. Volume distends an already failing right ventricle and drops cardiac output. Fluid can make the pressure worse in front of you.
This is permission to titrate instead of pour, not to skip resuscitation, and it comes with an obligation to document why.
Source control
The most underemphasized element in sepsis care, and the one residents leave off the plan. No antibiotic sterilizes an undrained abscess and no pressor fixes an obstructed infected kidney. If there is a collection, a device or dead tissue, it comes out or gets drained, and everything else is buying time until then.
- Abscess anywhere: intra-abdominal, hepatic, perinephric, epidural, dental, tubo-ovarian. Radiology drain or surgery.
- Empyema or complicated parapneumonic effusion. A loculated, low-pH, purulent effusion needs a chest tube, usually sooner than anyone wants.
- Obstructed ureter above an infection. Fever, flank pain, a stone and hydronephrosis is a urologic emergency. Nephrostomy or stent, tonight. This patient dies while you round.
- Cholangitis. Biliary decompression by ERCP, same urgency.
- Necrotizing soft tissue infection. Surgery. Not more imaging, not a consult in the morning. Pain out of proportion, rapid progression, crepitus, bullae or systemic toxicity means you call now.
- Infected line or device. Remove it. A peripheral or non-tunneled central line in a septic patient comes out.
- Perforated viscus, ischemic bowel, infected pancreatic necrosis, septic arthritis. Operative or interventional problems.
Identify the anatomic source and achieve definitive source control within 6 to 12 hours of recognition, and immediately in necrotizing infection. Every hour of delay is an hour your antibiotics are not working. If it is delayed for a defensible reason (too unstable to transport, awaiting interventional radiology, coagulopathy being corrected) write that reason with a time and a plan for when it happens.
Empiric antibiotics by suspected source
Common starting points for an adult with normal renal function. Doses change in renal failure, obesity and augmented renal clearance.
Local resistance decides whether cefepime is adequate empiric gram-negative coverage on your unit or whether you need a carbapenem. Institutional protocols, stewardship restrictions and pharmacy nomograms govern. Call infectious diseases at hour one rather than day five.
When to add MRSA coverage
Add it for a reason: known MRSA colonization or prior infection, a purulent skin or soft tissue source, a catheter as the suspected source, instability with an unclear source, recent hospitalization, injection drug use, necrotizing or post-influenza pneumonia, or high local prevalence. Not reflexively on every sepsis alert, and not left running for a week. Vancomycin 20 to 25 mg/kg actual weight as a load (commonly capped near 2 to 3 g), then 15 to 20 mg/kg every 8 to 12 hours with pharmacy adjusting to an AUC target. Linezolid 600 mg IV or PO every 12 hours is the alternative, preferred in acute kidney injury, watching serotonergic interactions and cytopenias after about two weeks.
When to add antipseudomonal coverage
Structural lung disease such as bronchiectasis or cystic fibrosis, intravenous antibiotics or hospitalization within 90 days, neutropenia, burns, prior Pseudomonas, or septic shock from an unknown source. Piperacillin-tazobactam plus vancomycin carries a consistent association with acute kidney injury; the mechanism is argued about, but many units now prefer cefepime plus vancomycin when the choice is otherwise a coin flip.
Ceftriaxone 1 to 2 g IV every 24 hours plus azithromycin 500 mg IV or PO daily, or levofloxacin 750 mg IV daily as monotherapy. Use ceftriaxone 2 g for severe disease. Substitute cefepime 2 g every 8 hours or piperacillin-tazobactam 4.5 g every 6 hours for pseudomonal risk, and add vancomycin or linezolid for MRSA risk.
Cefepime 2 g IV every 8 hours, piperacillin-tazobactam 4.5 g every 6 hours (extended infusion over 4 hours where pharmacy supports it), or meropenem 1 g every 8 hours, plus vancomycin or linezolid. A second antipseudomonal agent from a different class (tobramycin, or ciprofloxacin 400 mg IV every 8 hours) only in shock or high-resistance units, stopped the moment sensitivities return.
Community-onset pyelonephritis: ceftriaxone 1 g IV daily. Healthcare-associated, recently instrumented or catheterized: cefepime 2 g every 8 hours or piperacillin-tazobactam 4.5 g every 6 hours. Prior ESBL-producing organism: ertapenem 1 g IV daily, or meropenem 1 g every 8 hours if pseudomonal coverage is also needed. Ask about an obstructing stone or a retained stent.
Community-acquired: piperacillin-tazobactam 4.5 g every 6 hours, ceftriaxone 2 g daily plus metronidazole 500 mg IV every 8 hours, or ertapenem 1 g daily. Healthcare-associated or postoperative: cefepime plus metronidazole, or meropenem. Add micafungin 100 mg IV daily for upper gastrointestinal perforation, recurrent perforation, prolonged antibiotic exposure or known Candida colonization. Anaerobic coverage is mandatory, and metronidazole is the cheap way to get it if your beta-lactam does not.
Septic with nonpurulent cellulitis: vancomycin plus piperacillin-tazobactam, then narrow hard. Suspected necrotizing infection: vancomycin plus piperacillin-tazobactam plus clindamycin 900 mg IV every 8 hours, and a surgeon at the bedside. The clindamycin is for toxin suppression, not for extra gram-positive coverage.
Vancomycin plus an antipseudomonal gram-negative agent such as cefepime 2 g every 8 hours. Add an echinocandin for parenteral nutrition, a femoral line, prolonged broad-spectrum exposure or Candida colonization. Paired cultures from line and periphery, and take the line out.
Antibiotics within one hour, no waiting for imaging. Monotherapy with cefepime 2 g every 8 hours, piperacillin-tazobactam 4.5 g every 6 hours, or meropenem 1 g every 8 hours. Add vancomycin only for a specific reason: instability, catheter infection, skin or soft tissue source, radiographic pneumonia, known MRSA colonization, severe mucositis. Consider antifungal coverage for persistent fever after roughly 4 to 7 days in a still-neutropenic patient.
Ceftriaxone 2 g IV every 12 hours plus vancomycin. Add ampicillin 2 g IV every 4 hours for Listeria if the patient is over 50, pregnant, alcoholic or immunocompromised. Dexamethasone 0.15 mg/kg IV every 6 hours, starting with or just before the first dose, when pneumococcal meningitis is suspected. Add acyclovir 10 mg/kg IV every 8 hours for any encephalitic feature. Do not delay antibiotics for the CT or the tap. Cultures, drugs, then imaging.
Once you have an organism the job changes completely, which is the subject of the antibiotic duration and de-escalation page. Labeling for any individual agent is in the drug guide.
Vasopressors and steroids
The target is a MAP of 65 mmHg. Higher targets have not shown benefit in general septic populations and cost you arrhythmia and digital ischemia. A target of 75 to 80 is defensible in long-standing severe hypertension when the patient is not making urine at 65. Say why in the note.
Alpha-1 predominant with modest beta-1 activity, so it raises systemic vascular resistance without much tachycardia. Start around 0.05 mcg/kg/min and titrate. Nothing else seriously competes for this spot.
Fixed at 0.03 units/min, not titrated. It goes on as the norepinephrine dose climbs, commonly around 0.25 to 0.5 mcg/kg/min, though the trigger varies by unit. Two reasons it is second: septic shock produces a relative vasopressin deficiency, so you are replacing a missing hormone rather than pushing an exhausted receptor harder, and V1 receptors are not desensitized by the catecholamine storm. Wean the norepinephrine first; stopping vasopressin abruptly can drop the pressure sharply.
Third line, or second where vasopressin is unavailable. Useful when there is myocardial depression and you want inotropy with the vasoconstriction. It raises lactate through beta-2 mediated glycolysis, so a lactate that rises after you start it is not automatically worsening perfusion. Expect tachyarrhythmia.
Pure alpha-1, and not a septic shock drug. Narrow role: when tachyarrhythmia limits norepinephrine, in short procedural or induction hypotension, and as a push-dose bridge. Afterload with no inotropy can drop cardiac output in a failing heart.
Salvage for refractory vasodilatory shock on high-dose conventional agents. It reliably raises pressure and is catecholamine sparing; a mortality benefit is not established. Expensive, usually restricted, and thrombosis risk means protocols mandate venous thromboembolism prophylaxis. A conversation with critical care, not the next rung on your ladder.
200 mg per day, as 50 mg IV every 6 hours or a continuous infusion, for shock that stays refractory on adequate fluid and ongoing vasopressors. Some protocols add fludrocortisone 50 mcg daily. The evidence supports faster shock reversal and less pressor time; mortality data are genuinely mixed. Do not order a cosyntropin stimulation test to decide. Watch the glucose and taper rather than stopping abruptly after several days.
Access
Central access is right for anyone likely to need pressors beyond a few hours. An arterial line is right for anyone on a pressor at all, because a cuff pressure in a vasoconstricted patient is unreliable and you are titrating to a number you cannot trust. Both are on the arterial line and CVP page.
Waiting for a central line while the MAP sits at 50 is worse medicine than starting peripherally. Reasonable when the catheter is 20 gauge or larger, sited antecubital or more proximal (not hand, wrist or foot), with confirmed blood return, hourly site checks, and a limited duration while central access is arranged. Know where the phentolamine is before you need it. Your institutional policy may be stricter, and it wins.
Assessing the response
After the first bolus the question is not whether you completed the bundle. It is whether this patient is perfusing, and whether more fluid helps or hurts.
- Lactate clearance. Recheck at 2 to 4 hours. A flat or rising lactate on adequate resuscitation sends you back to two questions: is the source controlled, and is the antibiotic right. A stubborn lactate is far more often an undrained abscess or a wrong drug than a need for another liter.
- Capillary refill and peripheral perfusion. Press the finger pulp, release, count. Under 3 seconds is normal, and mottling over the knee is a bad sign. Free, repeatable, and resuscitation guided by capillary refill has been formally compared against lactate-guided resuscitation and performed at least as well.
- Urine output. Above roughly 0.5 mL/kg/hour. It lags and it is confounded, but it is a real end-organ signal.
- Mental status. The patient who starts talking to you again is responding.
Dynamic measures, not static numbers
This is about as settled as anything in critical care. A CVP of 5 does not mean a bolus will help and a 14 does not mean it will not. CVP reflects venous compliance, intrathoracic pressure, right ventricular function and tricuspid competence at least as much as volume. Stop writing "CVP 6, will give more fluid." If you use it at all, use the trend, and use it as a congestion signal rather than a filling target.
- Passive leg raise with a cardiac output measure. Start semi-recumbent at 45 degrees, then lay the torso flat and raise the legs to 45 degrees, autotransfusing roughly 300 mL. Measure stroke volume or velocity time integral before and about a minute after; a rise of roughly 10 percent or more predicts responsiveness. The detail people get wrong: it only works if you are measuring stroke volume. Watching the blood pressure alone is much less reliable.
- Pulse pressure or stroke volume variation. Valid only in a passive, sedated patient on controlled ventilation with adequate tidal volumes, in sinus rhythm, without right heart failure and without an open chest. On a spontaneously breathing floor patient in atrial fibrillation it means nothing.
- Bedside echocardiography. Ventricular size and function, pericardial effusion, right ventricular strain. A hyperdynamic underfilled ventricle and a severely depressed one call for opposite next moves.
- A fluid challenge with a defined endpoint. 250 to 500 mL over 10 to 15 minutes and a decision at the end of it. If nothing improved, the next bolus will not either.
Once shock resolves and the pressors are off, the plan becomes getting the fluid back off. Deresuscitation is an active phase of care and belongs in the plan by name.
Documenting sepsis
A sepsis chart is read later by three audiences: the next clinician, a coder, and possibly a plaintiff's attorney. All three want the same thing, evidence that you recognized this early, reasoned deliberately and acted. Give it to them.
Time-stamp the recognition
Write the moment you knew and what you knew. "At 14:20 the patient was febrile to 39.1, heart rate 122, blood pressure 84/48 with a MAP of 60, new confusion, lactate 4.1. I recognized sepsis with hypoperfusion at that time." Everything else is measured from that stamp. If you never write it, someone picks a time for you later, and it will not be the one you would have picked.
Say why you chose those antibiotics
Not "started vancomycin and cefepime." Write: "Empiric vancomycin and cefepime chosen for suspected hospital-acquired pneumonia given hospitalization 3 weeks ago, prior MRSA nares colonization in March, and a right lower lobe infiltrate. Will reassess at 48 hours with cultures." Eleven seconds, and it justifies a broad regimen to a stewardship reviewer while telling the day team what to reconsider.
Document the response to fluid, not just the volume
"Received 2 L lactated Ringer's" is data entry. "Received 2 L lactated Ringer's in 500 mL aliquots; MAP improved from 55 to 68 after the first liter, no further improvement with the second and no change in capillary refill, so fluid was held and norepinephrine started at 16:05" is medicine. The second shows reassessment after each intervention, which is the thing actually being judged.
Name the deviation and the reason
Bundles measure whether you did the thing, not whether the thing was right for this patient. Unexplained noncompliance reads as an oversight. Explained deviation reads as judgment.
The deviation sentence
"Full 30 mL/kg (2,250 mL) not given. Patient has an ejection fraction of 20 percent and presented with bibasilar crackles and a new oxygen requirement. Gave 1,000 mL in 250 mL aliquots with reassessment after each; MAP improved to 66 while the oxygen requirement rose from 2 to 4 L, so fluid was stopped and norepinephrine started. Risk of pulmonary edema judged to outweigh the benefit of further volume. Discussed with Dr. Rivera, intensivist, at 16:30."
Coding, briefly
Link the sepsis to the organism and the source: "sepsis due to Escherichia coli urinary tract infection" codes cleanly, and "urosepsis" does not code as sepsis at all. Name each organ dysfunction explicitly ("acute hypoxemic respiratory failure," "acute kidney injury," "acute metabolic encephalopathy") instead of leaving it implied in the labs. If it is septic shock, use the words septic shock, and if it was present on admission, say so. More on the billing and coding page; the principles of a chart that holds up are on the medical-legal charting page.
A worked note
Rapid response and sepsis evaluation note
Time of evaluation: 02:14. Reason: rapid response called by nursing for fever and hypotension.
Assessment. 68 year old man, hospital day 2 after right hip open reduction and internal fixation, with type 2 diabetes and chronic kidney disease stage 3. At 02:14 he is febrile to 38.9 C, heart rate 118, blood pressure 82/46 with a MAP of 58, respiratory rate 26, saturating 89 percent on room air and 94 percent on 4 L. Oriented to person only, from a fully oriented baseline. Extremities cool, capillary refill 4 seconds, mottling over both knees. Right basilar crackles. Surgical wound clean and dry. Foley in since surgery. Labs 02:20: white count 18.4 with 12 percent bands, platelets 96 (210 yesterday), creatinine 2.4 (baseline 1.3), lactate 4.6. Chest radiograph with a new right lower lobe opacity.
Impression. Sepsis with hypoperfusion, recognized 02:14, most likely from hospital-acquired right lower lobe pneumonia, with catheter-associated urinary tract infection a secondary consideration. Acute organ dysfunction in four systems: acute hypoxemic respiratory failure, acute kidney injury (creatinine 2.4 from 1.3), acute thrombocytopenia (96 from 210) and acute metabolic encephalopathy, an acute SOFA rise well over 2 points. Does not yet meet septic shock criteria, as no vasopressor has been given and resuscitation is ongoing.
Actions, with times.
02:18 Two sets of blood cultures from separate sites. Urine and sputum cultures sent. Foley removed and replaced.
02:26 Vancomycin 1,750 mg IV (25 mg/kg load, 70 kg) and cefepime 2 g IV given. Rationale: a new infiltrate on hospital day 2 meets hospital-acquired pneumonia criteria; antipseudomonal coverage for recent hospitalization, MRSA coverage for instability with a new infiltrate. Cefepime over piperacillin-tazobactam given the acute kidney injury and concurrent vancomycin. Pharmacy consulted for renal dosing.
02:30 Lactated Ringer's 500 mL over 15 minutes, repeated. Target 30 mL/kg equals 2,100 mL.
03:05 After 1,000 mL: MAP 58 to 63, capillary refill 3 seconds, oxygen requirement stable. Continued to 2,100 mL total by 03:40.
03:45 After the full volume: MAP 60, still cool peripherally, urine output 15 mL in the preceding hour, oxygen requirement now 6 L. Judged not fluid responsive; further bolus held.
03:50 Norepinephrine at 0.05 mcg/kg/min through a 20 gauge antecubital catheter with confirmed blood return, titrating to MAP 65, hourly site checks per the peripheral vasopressor policy, phentolamine on the unit. Central and arterial lines on ICU arrival.
04:00 Discussed with Dr. Okafor, intensivist. Accepted to the medical ICU.
Source control. No drainable collection identified. No effusion on chest radiograph; bedside ultrasound if fever or leukocytosis persists, to exclude empyema. Foley exchanged 02:18. No central catheter. No surgical site infection; orthopedics notified.
Reassessment plan. Repeat lactate 04:30. Hourly capillary refill and mental status while on the pressor. If the lactate is not clearing by 06:00, re-image for an undrained source and reconsider the antibiotic. Cultures reviewed at 48 hours with a documented de-escalation decision; MRSA nares PCR sent, and vancomycin stopped if it is negative with no gram-positive growth. Antibiotic review date: hospital day 4. Wife updated by telephone at 04:10 regarding the diagnosis, the ICU transfer and the possibility of intubation; code status confirmed full code.
Notice what that note does. It fixes a time, names each organ dysfunction, justifies every antibiotic, shows a reassessment after each intervention rather than a single volume total, addresses source control even though the answer was negative, and writes the de-escalation plan on day one. See also rapid response for the mechanics of the call itself.
This is not medical advice. It is a teaching outline for clinicians and clinicians in training. Every drug, dose and threshold below is illustrative. Your local antibiogram, your institution's sepsis protocol and a current reference govern what you actually order.